Target detection method based on background noise estimation self-updating

Abstract

The invention relates to a target detection method based on background noise estimation self-updating. The method comprises the following steps: judging whether a to-be-detected unit and a reference window around the to-be-detected unit belong to a homogeneous region or not; when the to-be-detected unit belongs to the mask, taking the background noise of the to-be-detected unit as updated background noise; when the to-be-detected unit does not belong to the mask, background noise of the to-be-detected unit is updated; calculating local noise estimation of the reference window, fusing the local noise estimation and the updated background noise, and calculating a fused noise estimation threshold; judging whether a target is detected in the to-be-detected unit according to the amplitude of the to-be-detected unit and the fusion noise estimation threshold; according to a detection result, calculating an actual false alarm rate, and according to a difference between the actual false alarm rate and a set false alarm rate, dynamically adjusting a fusion noise estimation threshold multiplying power to obtain an adjusted threshold multiplying power; and repeating the steps to adjust the actual false alarm rate until a preset condition is met, and obtaining a detection result. According to the method, target detection under a complex clutter background is realized.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of target detection, and particularly relates to a target detection method based on background noise estimation self-updating. BACKGROUND

[0002] The existing CFAR relies on a local reference window, and is prone to threshold mismatch, false alarm drift and missed detection in the face of non-homogeneous, strongly non-stationary clutter such as sea surface / terrain edge / multiple target dense; the clutter map provides a reliable reference for online detection by modeling the background power and distribution prior on a larger spatiotemporal scale, but the traditional approach only stores the mean / variance, which is easily contaminated by target energy, and is not adaptive to multi-dimensional scenes and waveform switching. Therefore, the technical development focuses on robustly fusing the clutter map prior with the current observation, combining homogeneity discrimination driven window shape / window length adaptation and sample weight reduction (or rejection), switching CA / OS / GO / SO strategies as needed and adapting non-Gaussian distribution parameters, while taking anti-pollution updates and dividing the graph by waveform / direction / sea state. The goal is to achieve long-term constant false alarm, rapid convergence in sudden scenes, and significantly improve detection probability and engineering maintainability in edge and strong scattering areas, with evaluation indicators such as false alarm stability, detection probability, convergence speed and algorithm latency.

[0003] A patent application with the application publication number CN114518564B and the name "a sea surface low-altitude small target detection method based on feature clutter map" discloses a sea surface low-altitude small target detection method based on feature clutter map. By extracting effective target and sea clutter distinguishing features, a feature clutter map under pure sea clutter background is established. The inter-frame accumulated clutter map is gradually stabilized, so that the discrimination between target and sea clutter is not affected by singular eigenvalues, the dimension is controllable, and the storage space occupied is small. In addition, the detection probability and false alarm probability are effectively controlled, the detection performance is guaranteed, and the target missed detection and false alarm problem is reduced. However, this method relies on the feature clutter map matrix under the pure sea clutter background. In actual application, especially under complex weather conditions or when the sea state changes, changes in the sea surface environment may affect the stability of the clutter map. For non-pure sea background (such as clutter with other objects or interference), the method may encounter difficulties, and although the method effectively controls the detection probability and false alarm probability, in complex background or dynamic changing scenes, especially when the difference between target and clutter is difficult to distinguish, false alarm and missed detection problems may still exist.

[0004] The patent application with the application publication number CN116559810A and the title of "An ordered clutter map constant false alarm rate detection algorithm" discloses an ordered clutter map constant false alarm rate detection algorithm, which relates to the technical field of radar detection algorithms. First, the OS-CFAR algorithm is used to process the radar scanning data to obtain a to-be-detected cell. Then, the CM-CFAR algorithm is used to process to obtain a background clutter power level estimation, and a detection threshold is calculated using a nominal factor. Finally, the detection threshold is compared with the amplitude of the to-be-detected cell to complete the constant false alarm rate detection. Although this method can achieve good results in uniform environments, multi-target environments, and clutter edge environments, the performance of the algorithm may be affected in more complex clutter backgrounds (for example, rapidly changing environments, non-uniform clutter, or dynamic interference environments). The dynamic characteristics of the clutter may cause inaccurate background estimation, which in turn affects the detection results. Moreover, the algorithm involves two different CFAR algorithms (OS-CFAR and CM-CFAR), as well as the calculation and comparison of the detection threshold, which may result in high computational complexity of the algorithm. In applications requiring high real-time performance, the computational burden may become a limiting factor.

[0005] Therefore, existing constant false alarm rate target detection algorithms are difficult to solve target detection in complex clutter backgrounds, and have high computational complexity. SUMMARY

[0006] To solve the above problems in the prior art, the present application provides a target detection method based on background noise estimation self-update. The technical problem to be solved by the present application is solved by the following technical scheme: The present application provides a target detection method based on background noise estimation self-update, comprising the following steps: S1, determining whether the to-be-detected cell and its surrounding reference window belong to the same homogeneous region to obtain a determination result; S2, when the to-be-detected cell belongs to a mask, the background noise estimation of the to-be-detected cell is taken as the updated background noise estimation; when the to-be-detected cell does not belong to the mask, the amplitude mean of the reference cell is updated based on the determination result to update the background noise estimation of the to-be-detected cell, and the updated background noise estimation is obtained; S3, calculating the local noise estimation of the reference window based on the determination result, fusing the local noise estimation and the updated background noise estimation, and calculating the fused noise estimation threshold using the fused value; S4, judging whether a target is detected in the to-be-detected cell according to the amplitude of the to-be-detected cell and the fused noise estimation threshold to obtain a detection result; S5, calculating the actual false alarm rate according to the detection result, and dynamically adjusting the fused noise estimation threshold multiple according to the difference between the actual false alarm rate and the set false alarm rate to obtain an adjusted threshold multiple. S6. Based on the adjusted threshold multiplier, repeat S3-S5 to adjust the actual false alarm rate until the difference between the actual false alarm rate and the set false alarm rate meets the preset condition, and obtain the detection result.

[0007] In one embodiment of the present invention, step S1 includes: Select the unit to be detected and a reference window around it, and calculate the variation index based on the amplitude value of the reference window:

[0008] in, The variation index, The number of reference windows around the unit to be detected. For the first The amplitude value of each reference window; When the variation index is determined to be greater than a preset threshold, the unit to be detected and the surrounding reference windows belong to a homogeneous region. When the variation index is determined to be less than or equal to the preset threshold, the unit to be detected and the surrounding reference windows belong to a non-homogeneous region.

[0009] In one embodiment of the present invention, the updated background noise estimate in step S2 is:

[0010] in, For the updated number The first frame Background noise estimation for each unit to be detected For the first The first frame Background noise estimation for each unit to be detected These are the weighting coefficients; The average amplitude of all reference windows is used as the reference value. When the unit to be detected and the surrounding reference windows belong to the same region, the average amplitude of all reference windows is used as the reference value. When the unit to be detected and the surrounding reference windows belong to a non-homogeneous region, the median or truncated mean of the amplitudes of all reference windows is used as... .

[0011] In one embodiment of the present invention, step S3, which calculates the local noise estimate of the reference window based on the judgment result, includes: When the unit to be detected and the surrounding reference window belong to the same region, the local noise estimate of the reference window is calculated using the CA-CFAR method:

[0012] When the unit to be detected and the surrounding reference window belong to a non-homogeneous region, the local noise estimate of the reference window is calculated using the OS-CFAR method:

[0013] in, For local noise estimation of the reference window, For the first The amplitude value of each reference window, The number of reference windows around the unit to be detected. This is the k-th value taken when the reference window is sorted by amplitude from smallest to largest.

[0014] In one embodiment of the present invention, the fused value in step S3 is:

[0015] in, This is the value obtained by fusing the local noise estimate and the background noise estimate. To estimate update coefficients for background noise, For the updated number The first frame Background noise estimation for each unit to be tested.

[0016] In one embodiment of the present invention, the fusion noise estimation threshold in step S3 is:

[0017]

[0018] in, This is the value obtained by fusing the local noise estimate and the background noise estimate. To estimate update coefficients for background noise, For the updated number The first frame Background noise estimation for each unit to be detected To estimate the threshold for fusion noise, For the first Frame threshold ratio, The threshold multiplier for the first frame. This is the set constant false alarm probability.

[0019] In one embodiment of the present invention, step S4 includes: When the amplitude of the unit to be detected is greater than the fusion noise estimation threshold, a target is detected in the unit to be detected, and a mask is generated between the unit to be detected and the surrounding reference window; When the amplitude of the unit to be detected is less than or equal to the fusion noise estimation threshold, no target is detected in the unit to be detected, and the unit to be detected is marked as background.

[0020] In one embodiment of the present invention, step S5 includes: Calculate the actual false alarm rate based on the detection results:

[0021] The difference between the actual false alarm rate and the set false alarm rate is determined, and the fusion noise estimation threshold ratio is adjusted according to the magnitude of the difference to obtain the adjusted threshold ratio:

[0022] in, It is the adjusted number Frame threshold ratio, It is the first before the adjustment Frame threshold ratio, For the feedback gain of the self-calibration loop, This represents the actual false alarm rate. To set the false alarm rate. In one embodiment of the present invention, the preset condition in step S6 is: ,in This represents the actual false alarm rate. To set the false alarm rate. Compared with the prior art, the beneficial effects of the present invention are as follows: 1. In the method of the present invention, it is first determined whether the unit to be detected and the surrounding reference window belong to the same region. Based on the determination result, the background noise estimate is updated and the local noise estimate of the reference window is calculated. Then, the local noise estimate and the updated background noise estimate are fused to realize robust noise estimation and CFAR threshold generation in heterogeneous clutter environment. 2. In the method of the present invention, when the unit to be detected belongs to the mask, the background noise estimate of the unit to be detected is used as the updated background noise estimate, freezing the target area and preventing the target from backfeeding the background noise estimate. At the same time, it is determined whether the target is detected in the unit to be detected based on the amplitude of the unit to be detected and the fusion noise estimation threshold, so as to distinguish the target from the background and update the mask to protect the target signal, prevent the background noise estimate update from contaminating the target energy, and improve the detection rate of small targets. 3. Compared with traditional CFAR and other algorithms, the method of this invention solves the problems of heterogeneous clutter, target absorption, false alarm rate fluctuation and insufficient multidimensional detection accuracy in the existing technology, realizes target detection in complex clutter background, and can be adaptively applied to different CFAR detection methods for target detection. Attached Figure Description

[0023] Figure 1 A flowchart illustrating a target detection method based on background noise estimation and self-updating, provided in an embodiment of the present invention; Figure 2 A flowchart illustrating another target detection method based on background noise estimation self-updating provided in an embodiment of the present invention; Figure 3 The figure shows the Pd-Pfa simulation performance of the CA-CFAR algorithm under complex interference conditions with signal-to-noise ratios of 9dB, 12dB, and 15dB. Figure 4 The following are simulation results of the OS-CFAR algorithm under complex interference conditions, with Pd-Pfa parameters at signal-to-noise ratios of 9dB, 12dB, and 15dB. Figure 5 The simulation performance of the method of the present invention under complex interference conditions is shown in the Pd-Pfa simulation diagrams at signal-to-noise ratios of 9dB, 12dB, and 15dB. Figure 6 A comparison of the detection probability-signal-clutter ratio curves of CA-CFAR, OS-CFAR, and the method of this patent under complex interference conditions; Figure 7 The CA-CFAR algorithm is used to detect targets for five targets under complex interference conditions with signal-to-noise ratios of 9dB, 12dB, and 15dB. Figure 8 The OS-CFAR algorithm provides target detection images for five targets under complex interference conditions with signal-to-noise ratios of 9dB, 12dB, and 15dB. Figure 9 For target detection maps of 5 targets under complex interference conditions, the method of the present invention is used under signal-to-noise ratios of 9dB, 12dB, and 15dB. Detailed Implementation

[0024] The present invention will be further described in detail below with reference to specific embodiments, but the implementation of the present invention is not limited thereto.

[0025] Example 1 Please see Figure 1 and Figure 2 , Figure 1 This is a flowchart illustrating a target detection method based on background noise estimation and self-updating, provided in an embodiment of the present invention. Figure 2 This is a flowchart illustrating another target detection method based on background noise estimation self-updating, provided in an embodiment of the present invention.

[0026] This embodiment addresses the problems of low detection rate and poor detection performance of traditional target detection algorithms in complex clutter backgrounds, and proposes a target detection method based on background noise estimation and self-updating, including the following steps: S1. Determine whether the cell under test (CUT) and the surrounding reference windows belong to the same region, and obtain the determination result.

[0027] Specifically, the first step is to select the cell to be detected and its surrounding reference windows. Taking a 5×5 cell as an example, if the center cell is selected as the cell to be detected, then the cells surrounding the center cell can be used as reference windows, or the two rings of cells surrounding the center cell can be used as reference windows. In this embodiment, the selection of the cell to be detected and its surrounding reference windows is not limited to these.

[0028] Then, the variation index is calculated based on the amplitude value of the reference window:

[0029] in, The variation index, The number of reference windows around the unit to be detected. For the first The amplitude value of each reference window.

[0030] Finally, non-homogeneity detection (NHD) is performed. When the variance index is greater than a preset threshold, the unit under test and the surrounding reference windows belong to a homogeneous region. When the variance index is less than or equal to the preset threshold, the unit under test and the surrounding reference windows belong to a non-homogeneous region. A homogeneous region refers to a region with uniform noise, while a non-homogeneous region refers to a region with spike interference or a target area. For example, the preset threshold is empirically set to 2.

[0031] S2. When the unit to be detected belongs to the mask, the background noise estimate of the unit to be detected is used as the updated background noise estimate, and the target area is frozen to avoid the target backfeeding the background noise estimate. When the unit to be detected does not belong to the mask, the amplitude mean of the reference unit is updated based on the judgment result to update the background noise estimate of the unit to be detected, and the long-term background noise estimate is updated to obtain the updated background noise estimate, thereby reducing the influence of heterogeneous clutter on the threshold.

[0032] Specifically, the updated background noise estimate is as follows:

[0033] in, For the updated number The first frame Background noise estimation for each unit to be detected For the first The first frame Background noise estimation for each unit to be detected These are the weighting coefficients; The average amplitude of the reference cells is used as the mean amplitude. When the cell to be detected and the surrounding reference windows belong to the same region, the average amplitude of all reference windows is used as the mean amplitude. When the unit to be detected and the surrounding reference windows belong to a non-homogeneous region, the median or truncated mean of the amplitudes of all reference windows is used as the mean. .

[0034] S3. Based on the judgment result, calculate the local noise estimate of the reference window around the unit to be detected, fuse the local noise estimate and the updated background noise estimate, and use the fused value to calculate the fused noise estimate threshold.

[0035] Specifically, firstly, when the cell to be detected and the surrounding reference window belong to the same region, the local noise estimate of the reference window is calculated using the CA-CFAR (Cell Averaging CFAR) method:

[0036] When the cell to be detected and the surrounding reference window belong to a non-homogeneous region, the local noise estimate of the reference window is calculated using the OS-CFAR (OrderStatistics CFAR) method:

[0037] in, For local noise estimation of the reference window, For the first The amplitude value of each reference window, The number of reference windows around the unit to be detected. This is the k-th value taken when the reference window is sorted by amplitude from smallest to largest.

[0038] Then, the local noise estimate and the updated background noise estimate are fused to obtain the fused value:

[0039] in, This is the value obtained by fusing the local noise estimate and the background noise estimate. To estimate update coefficients for background noise, For the updated number The first frame Background noise estimation for each unit to be tested.

[0040] Finally, the fused values ​​are used to generate a robust fusion noise estimation threshold:

[0041]

[0042] in, This is the value obtained by fusing the local noise estimate and the background noise estimate. To estimate update coefficients for background noise, For the updated number The first frame Background noise estimation for each unit to be detected To estimate the threshold for fusion noise, For the first The frame threshold ratio is also known as the threshold factor. This is the threshold multiplier for the first frame, i.e., the initial threshold multiplier value. This is the set constant false alarm probability.

[0043] S4. Determine whether the target is detected in the unit to be detected based on the amplitude of the unit to be detected and the fusion noise estimation threshold, and obtain the detection result.

[0044] Specifically, based on the amplitude of the input detection unit The system also includes a fusion noise estimation threshold T to determine whether a target is detected. Undetected units are marked as background, while detected units are marked as having a target. A mask is generated for the detected units and the reference windows around them for background noise estimation updates in the next frame.

[0045] Furthermore, when the amplitude of the unit to be detected is greater than the fusion noise estimation threshold, a target is detected in the unit to be detected, i.e. The detection unit and the surrounding reference windows are used to generate a mask; when the amplitude of the detection unit is less than or equal to the fusion noise estimation threshold, no target is detected in the detection unit, and the detection unit is marked as background.

[0046] In this embodiment, the threshold generated by the fusion of background noise estimation and local noise estimation is applied to the unit to be detected to determine whether there is a target in the unit to be detected, thereby distinguishing the target from the background and updating the mask to protect the target signal and avoid background noise estimation contamination.

[0047] S5. Calculate the actual false alarm rate based on the detection results, and dynamically adjust the fusion noise estimation threshold multiple based on the difference between the actual false alarm rate and the set false alarm rate to obtain the adjusted threshold multiple. The self-calibration loop adjusts the threshold factor proportionally based on the difference between the actual false alarm rate and the set false alarm rate, thereby adjusting the threshold and making the actual false alarm rate close to the set false alarm rate, i.e., self-calibration.

[0048] First, calculate the actual false alarm rate based on the detection results of step S4.

[0049] In some scenarios, we have prior knowledge of the radar monitoring area, such as the absence of targets in certain azimuth or range segments, or certain areas serving as "background learning zones" specifically for estimating the false alarm rate. After CFAR detection, only the number of units identified as targets in these "target-free areas" is counted.

[0050] Then, the difference between the actual false alarm rate and the set false alarm rate is determined, and the fusion noise estimation threshold ratio is adjusted according to the magnitude of the difference to obtain the adjusted threshold ratio:

[0051] in, It is the adjusted number The frame threshold ratio is also known as the threshold factor. It is the first before the adjustment Frame threshold multiplier; The feedback gain of the self-calibration loop is typically 0.05~0.1. Large adjustments can be made quickly, but they are prone to oscillations. Small, converges smoothly, but responds slowly; This represents the actual false alarm rate. To set the false alarm rate. This embodiment maintains a constant false alarm rate for CFAR detection and dynamically adjusts the threshold multiplier. This enhances the robustness of detection under heterogeneous clutter, dynamic changes in the target, or fluctuations in background noise.

[0052] S6. Based on the adjusted threshold ratio, repeat S3-S5 to adjust the actual false alarm rate until the difference between the actual false alarm rate and the set false alarm rate meets the preset condition, and then obtain the detection result.

[0053] Specifically, the adjusted threshold multiplier is substituted into the noise estimation threshold in step S3. In the calculation formula, and repeat steps S3-S5 until the difference between the actual false alarm rate and the set false alarm rate meets the requirement. The test results were obtained at that time.

[0054] This embodiment further verifies the effectiveness of the above method through simulation.

[0055] Please see Figure 3 , Figure 4 , Figure 5 , Figure 3 The following are the Pd-Pfa simulation performance (ROC plot) of the CA-CFAR algorithm under complex interference conditions with signal-to-noise ratios of 9dB, 12dB, and 15dB. Figure 4The figures show the Pd-Pfa simulation performance of the OS-CFAR algorithm under complex interference conditions with signal-to-noise ratios of 9dB, 12dB, and 15dB. Figure 5 The simulation performance of the method of this invention under complex interference conditions, with signal-to-noise ratios of 9dB, 12dB, and 15dB, is shown in the Pd-Pfa simulation diagrams. Figure 3 , 4 As can be seen from 5, the method of the present invention has better detection sensitivity and anti-false alarm performance compared with traditional target detection algorithms such as CA-CFAR and OS-CFAR.

[0056] Please see Figure 6 , Figure 6 The figure shows a comparison of the detection probability-signal-to-clutter ratio (Pd-SCR) curves of CA-CFAR, OS-CFAR, and the method of this patent under complex interference conditions. Pd-SCR is one of the most intuitive indicators of the actual detection performance of the CFAR detection algorithm. As can be seen from the figure, the method of this invention has better sensitivity and anti-clutter capability compared to traditional target detection algorithms such as CA-CFAR and OS-CFAR.

[0057] Please see Figure 7 , Figure 8 , Figure 9 , Figure 7 The CA-CFAR algorithm is used to detect target images of five targets under complex interference conditions with signal-to-noise ratios of 9dB, 12dB, and 15dB. Figure 8 The OS-CFAR algorithm is used to detect target images of five targets under complex interference conditions with signal-to-noise ratios of 9dB, 12dB, and 15dB. Figure 9 To address complex interference conditions, the method of this invention provides target detection maps for five targets at signal-to-noise ratios of 9dB, 12dB, and 15dB. Figure 7 , 8 As can be seen from 9, the method of the present invention has better target detection performance than traditional target detection algorithms such as CA-CFAR and OS-CFAR.

[0058] In this embodiment, the method first determines whether the unit to be detected and the surrounding reference window belong to the same region. Based on the determination result, the background noise estimate is updated, and the local noise estimate of the reference window is calculated. Then, the local noise estimate and the updated background noise estimate are fused to achieve robust noise estimation and CFAR threshold generation in heterogeneous clutter environments.

[0059] In this embodiment, when the unit to be detected is part of the mask, the background noise estimate of the unit to be detected is used as the updated background noise estimate, freezing the target area and preventing the target from being fed back into the background noise estimate. At the same time, the amplitude of the unit to be detected and the fusion noise estimation threshold are used to determine whether the target is detected in the unit to be detected, thereby distinguishing the target from the background and updating the mask to protect the target signal, preventing the background noise estimate update from contaminating the target energy, and improving the detection rate of small targets.

[0060] This embodiment of the method solves the problems of heterogeneous clutter, target absorption, false alarm rate fluctuation and insufficient multidimensional detection accuracy in the prior art through steps such as background noise estimation construction, non-uniformity detection (NHD), fusion of local noise and background noise estimation, intelligent CFAR switching, and threshold self-calibration. It has the characteristics of high robustness in heterogeneous clutter environment, strong protection capability for small targets and excellent adaptability to dynamic environment, realizing target detection in complex clutter background and can be adaptively applied to different CFAR detection methods for target detection.

[0061] The above description, in conjunction with specific preferred embodiments, provides a further detailed explanation of the present invention. It should not be construed that the specific implementation of the present invention is limited to these descriptions. For those skilled in the art, various simple deductions or substitutions can be made without departing from the concept of the present invention, and all such modifications and substitutions should be considered within the scope of protection of the present invention.

Claims

1. A target detection method based on background noise estimation and self-updating, characterized in that, Including the following steps: S1. Determine whether the unit to be detected and the surrounding reference window belong to the same region, and obtain the judgment result; S2. When the unit to be detected belongs to the mask, the background noise estimate of the unit to be detected is used as the updated background noise estimate; when the unit to be detected does not belong to the mask, the amplitude mean of the reference unit is updated based on the judgment result to update the background noise estimate of the unit to be detected, thus obtaining the updated background noise estimate. S3. Calculate the local noise estimate of the reference window based on the judgment result, fuse the local noise estimate and the updated background noise estimate, and use the fused value to calculate the fused noise estimate threshold; S4. Determine whether a target is detected in the unit to be detected based on the amplitude of the unit to be detected and the fusion noise estimation threshold, and obtain the detection result; S5. Calculate the actual false alarm rate based on the detection results, and dynamically adjust the fusion noise estimation threshold ratio based on the difference between the actual false alarm rate and the set false alarm rate to obtain the adjusted threshold ratio. S6. Based on the adjusted threshold multiplier, repeat S3-S5 to adjust the actual false alarm rate until the difference between the actual false alarm rate and the set false alarm rate meets the preset condition, and then obtain the detection result.

2. The target detection method based on background noise estimation and self-updating according to claim 1, characterized in that, Step S1 includes: Select the unit to be detected and a reference window around it, and calculate the variation index based on the amplitude value of the reference window: in, The variation index, The number of reference windows around the unit to be detected. For the first The amplitude value of each reference window; When the variation index is determined to be greater than a preset threshold, the unit to be detected and the surrounding reference windows belong to a homogeneous region. When the variation index is determined to be less than or equal to the preset threshold, the unit to be detected and the surrounding reference windows belong to a non-homogeneous region.

3. The target detection method based on background noise estimation and self-updating according to claim 1, characterized in that, The updated background noise estimate in step S2 is: in, For the updated number The first frame Background noise estimation for each unit to be detected For the first The first frame Background noise estimation for each unit to be detected These are the weighting coefficients; The average amplitude of all reference windows is used as the reference value. When the unit to be detected and the surrounding reference windows belong to the same region, the average amplitude of all reference windows is used as the reference value. When the unit to be detected and the surrounding reference windows belong to a non-homogeneous region, the median or truncated mean of the amplitudes of all reference windows is used as... .

4. The target detection method based on background noise estimation and self-updating according to claim 1, characterized in that, Step S3, which calculates the local noise estimate of the reference window based on the judgment result, includes: When the unit to be detected and the surrounding reference window belong to the same region, the local noise estimate of the reference window is calculated using the CA-CFAR method: When the unit to be detected and the surrounding reference window belong to a non-homogeneous region, the local noise estimate of the reference window is calculated using the OS-CFAR method: in, For local noise estimation of the reference window, For the first The amplitude value of each reference window, The number of reference windows around the unit to be detected. This is the k-th value taken when the reference window is sorted by amplitude from smallest to largest.

5. The target detection method based on background noise estimation self-updating according to claim 4, characterized in that, The fused value described in step S3 is: in, This is the value obtained by fusing the local noise estimate and the background noise estimate. To estimate update coefficients for background noise, For the updated number The first frame Background noise estimation for each unit to be tested.

6. The target detection method based on background noise estimation self-updating according to claim 5, characterized in that, The fusion noise estimation threshold mentioned in step S3 is: in, This is the value obtained by fusing the local noise estimate and the background noise estimate. To estimate update coefficients for background noise, For the updated number The first frame Background noise estimation for each unit to be detected To estimate the threshold for fusion noise, For the first Frame threshold ratio, The threshold multiplier for the first frame. This is the set constant false alarm probability.

7. The target detection method based on background noise estimation self-updating according to claim 1, characterized in that, Step S4 includes: When the amplitude of the unit to be detected is greater than the fusion noise estimation threshold, a target is detected in the unit to be detected, and a mask is generated between the unit to be detected and the surrounding reference window; When the amplitude of the unit to be detected is less than or equal to the fusion noise estimation threshold, no target is detected in the unit to be detected, and the unit to be detected is marked as background.

8. The target detection method based on background noise estimation self-updating according to claim 1, characterized in that, Step S5 includes: Calculate the actual false alarm rate based on the detection results: The difference between the actual false alarm rate and the set false alarm rate is determined, and the fusion noise estimation threshold ratio is adjusted according to the magnitude of the difference to obtain the adjusted threshold ratio: in, It is the adjusted number Frame threshold ratio, It is the first before the adjustment Frame threshold ratio, For the feedback gain of the self-calibration loop, This represents the actual false alarm rate. To set the false alarm rate.

9. The target detection method based on background noise estimation and self-updating according to claim 1, characterized in that, The preset conditions in step S6 are: ,in This represents the actual false alarm rate. To set the false alarm rate.

Images